TeslaThe Texas Gigafactory is expanding to accommodate aSupercomputer cluster. Liang Jian, CEO of Supermicro, later posted a message on X to congratulate Musk, saying that its large-scale AI data center uses Supermicro's liquid cooling technology, which will help protect 20 billion trees for the earth.
As we all know, AI data centers often mean huge energy consumption, and Super Micro Computer hopes to reduce this pressure by promoting its liquid cooling technology. According to reports, compared with air cooling, this liquid cooling solution can reduce the electricity cost of the cooling infrastructure by 89%.
Liang said that AMD's goal is to "increase the adoption rate of liquid cooling solutions from less than 1% to more than 30% within a year." He estimates that liquid cooling solutions can save data centers $20 billion (currently about 145.643 billion yuan) in energy costs each year.
Musk is massively expanding the AI facilities at Tesla's Texas Gigafactory, even building a super-large "fan". The new expansion can accommodate 50,000 Nvidia GPUs and more Tesla's self-developed AI hardware, which is designed to be used to train Tesla's FSD system.
Musk estimates that the supercomputer cluster will consume 130 megawatts of electricity per year, and is expected to grow to more than 500 megawatts in 18 months. Musk claims that the construction of the facility is nearing completion and is scheduled to be put into use in the next few months.
It is important to note that Musk actually built not only the Tesla Gigafactory AI supercomputer cluster, but also xA Built a multi-billion dollar supercomputer cluster.
Strictly speaking, the xAI supercomputer cluster is more famous than the Tesla one. Musk has previously confirmed the purchase of 100,000 Nvidia H100 GPUs, which xAI will use to train its Grok model.
According to Musk, the xAI supercomputer cluster is expected to be completed "within a few months" and will also use the liquid cooling system of AMD. It is planned to upgrade 300,000 Nvidia B200 GPUs next summer. In other words, xAI has a higher priority than Tesla, so Musk asked Nvidia in June to ship thousands of GPUs originally supplied to Tesla to xAI. This move indirectly led to the delay of Tesla's supercomputer cluster for several months.